LW-DETR: A Transformer Replacement to YOLO for Real-Time Detection

Open paper
Benchmark Model Rank Results
real-time-object-detection-on-coco-val2017LW-DETR-xlarge (Objects365 pre-training)#6box AP: 58.3FPS (batch 1, GPU in brackets): 52 (T4)
real-time-object-detection-on-coco-val2017LW-DETR-large (Objects365 pre-training)#14box AP: 56.1FPS (batch 1, GPU in brackets): 114 (T4)
real-time-object-detection-on-coco-val2017LW-DETR-medium (Objects365 pre-training)#58box AP: 52.5FPS (batch 1, GPU in brackets): 179 (T4)
real-time-object-detection-on-coco-val2017LW-DETR-small (Objects365 pre-training)#94box AP: 48.0FPS (batch 1, GPU in brackets): 345 (T4)
real-time-object-detection-on-coco-val2017LW-DETR-tiny (Objects365 pre-training)#117box AP: 42.6FPS (batch 1, GPU in brackets): 500 (T4)

Rank counts only results with a code link.